crm-digital-fte-factory

Format agent responses according to channel-specific style guidelines before sending to customers. This skill should be used AFTER escalation-decision returns should_escalate=False and BEFORE calling send_response tool. Transforms raw response text into channel-appropriate format (Gmail: formal & thorough, WhatsApp: casual & rapid under 60 words, Web Form: direct & functional). Applies CloudStream CRM brand voice: empathetic, concise, proactive. Mandatory for EVERY customer response.

Identify or create customers based on email/phone and retrieve their history. This skill should be used on EVERY incoming message before any other processing. Extracts contact info from message metadata, resolves to unified customer_id, and loads interaction history.

Determine whether a customer interaction requires human escalation based on sentiment trends, conversation context, knowledge base failures, and business-specific triggers. This skill should be used AFTER generating a response but BEFORE sending it to the customer. Analyzes mandatory criteria (sentiment floor, negative trend, looping, explicit requests, high-value accounts) and business-specific triggers (legal threats, security breaches, billing issues, churn risk, technical complexity). Outputs structured escalation decision with priority level.

Retrieve relevant product documentation and knowledge base articles using semantic search. Use this whenever a customer asks product-related questions, "how-to" guides, or technical specs. This skill leverages the search_knowledge_base tool from the crm-digital-fte MCP server.

Analyze sentiment of customer messages in real-time to detect frustration, anger, or dissatisfaction. MANDATORY for EVERY incoming customer message from any channel (Gmail, WhatsApp, Web Form). Must be executed BEFORE any other processing or response generation. Critical for early detection of at-risk customers requiring immediate escalation.

Build production-ready Model Context Protocol (MCP) servers using the official MCP SDK. Scaffold, configure, secure, and deploy MCP servers with best practices for security, monitoring, and reliability. Use for creating robust MCP servers that integrate with Claude Code.

Connect Model Context Protocol (MCP) servers with OpenAI Agents SDK using various transport methods including Hosted MCP, Streamable HTTP, HTTP with SSE, and stdio. Provides guidance for selecting appropriate transport methods, configuring connections with proper authentication, implementing error handling, and following best practices for security and performance. Use when connecting MCP servers to AI applications, configuring transport protocols, setting up authentication, or troubleshooting connection issues.

Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations.